Enhancing image based classification for crop disease detection using a multiclass SVM approach with kernel comparison.

Agricultural production is still quite susceptible to plant diseases, despite the fact that it is essential to both economic growth and food security. Yellow rust can lower wheat yields by 20-30%, red rust by 5-10%, and anthracnose by up to 60% in crops including cotton and mango. For losses to be minimized, early and precise detection is therefore crucial. Preprocessing, segmentation, feature extraction, and classification are all included in this study's machine learning-based framework for detecting various crop leaf diseases. To test the model, 9,111 carefully chosen images that were balanced through augmentation were employed. The novelty of this work lies in combining bilateral filtering and GraphCut segmentation with texture-based feature extraction and a systematic comparison of multiclass SVM kernels across a multi-crop dataset. Experimental results using stratified 5-fold cross-validation show that the linear kernel SVM achieved the best performance, with 99.0% accuracy, 98.6% precision, 98.7% recall, and 98.6% F1-score-outperforming earlier SVM-based approaches. These findings demonstrate the effectiveness of kernel selection and preprocessing in enhancing disease classification and provide a strong basis for future comparisons with deep learning methods to build scalable and reliable plant disease detection systems.